pith. sign in

arxiv: 2604.22765 · v1 · submitted 2026-03-24 · 💻 cs.CY · cs.AI

Algorithmic Administration and the EU AI Act: Legal Principles for Public Sector Use of AI

Georgios Pavlidis , Ioannis Kastanas This is my paper

Pith reviewed 2026-05-15 01:06 UTC · model grok-4.3

classification 💻 cs.CY cs.AI
keywords EU AI Actadministrative lawpublic sector AIproportionalityaccountabilityalgorithmic administrationtransparencyreviewability
0
0 comments X

The pith

The EU AI Act's risk-based rules fall short of administrative law's proportionality principle for public sector AI and need extra safeguards.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines how the EU AI Act applies to public authorities using AI for decisions in areas such as social benefits, migration, education, and law enforcement. It finds that the Act's obligations on high-risk systems do not fully align with core administrative principles including discretion, the duty to state reasons, and proportionality. This misalignment risks undermining accountability, transparency, and the ability to review automated decisions. A sympathetic reader would care because government AI affects individual rights and the fairness of public administration on a large scale.

Core claim

The EU AI Act imposes transparency, oversight, and conformity requirements on high-risk AI deployed by public authorities, yet its risk-based classification does not match the administrative law principle of proportionality. As a result, interpretative strategies and additional safeguards are required to secure accountability, transparency, and reviewability when public decisions are made by automated systems.

What carries the argument

The EU AI Act's risk-based approach, which assigns obligations according to defined risk levels for AI systems in public administration and interacts with principles of administrative discretion, duty to state reasons, and proportionality.

If this is right

  • Public sector deployers of high-risk AI must adopt targeted safeguards beyond the Act's minimum rules to meet proportionality.
  • Interpretative strategies are needed to adapt the duty to state reasons and reviewability for opaque AI outputs.
  • Accountability mechanisms in domains such as migration and social benefits require explicit alignment with administrative discretion rules.
  • The Act alone does not guarantee the rule of law in automated public decision-making without further legal adaptation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Member states may need to develop national guidelines that supplement the AI Act with AI-specific proportionality tests.
  • Courts could be asked to create new standards for reviewing algorithmic decisions that treat AI as distinct from human discretion.
  • Pilot implementations in EU agencies could test whether current risk tiers produce proportionate outcomes in practice.

Load-bearing premise

Traditional administrative law principles developed for human decision-makers apply directly to AI systems without needing fundamental modification.

What would settle it

A documented case in which a public authority fully complies with the AI Act for a high-risk system in law enforcement or benefits administration and the automated decision is still found proportionate and reviewable under existing administrative law standards.

read the original abstract

The increasing use of artificial intelligence (AI) by public authorities introduces both opportunities for innovation and significant challenges for the administrative rule of law. This article examines how the EU AI Act interacts with the fundamental principles of administrative law, with a particular focus on administrative discretion, the duty to state reasons, and proportionality. It analyses the regulatory obligations imposed by the AI Act on public sector deployers of high-risk systems, especially in sensitive domains such as social benefits, migration, education, and law enforcement. It also explores whether the AI Act adequately ensures accountability, transparency, and reviewability in automated public decision-making. The article further considers how the AI Act's risk-based approach aligns (or fails to align) with the principle of proportionality and it proposes safeguards and interpretative strategies to ensure the ethical and lawful deployment of AI in the public sector.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The paper claims that the increasing deployment of AI by public authorities challenges the administrative rule of law and that the EU AI Act's risk-based framework interacts with core administrative-law principles (discretion, duty to state reasons, proportionality). It analyzes obligations imposed on public-sector deployers of high-risk systems in domains such as social benefits, migration, education and law enforcement, argues that the Act's risk classification does not fully align with proportionality, and proposes interpretative strategies and safeguards to secure accountability, transparency and reviewability in automated public decisions.

Significance. If the doctrinal mapping holds, the paper supplies a timely bridge between the EU AI Act and established administrative-law doctrines, offering concrete guidance for public authorities on how to interpret and operationalise the Act's obligations. By identifying gaps in proportionality and reviewability and suggesting targeted safeguards, it contributes to the practical implementation of lawful AI use in sensitive public-sector contexts and may inform both legal scholarship and regulatory practice.

major comments (1)
  1. [Section analysing proportionality and risk classification] The central claim that the AI Act's risk-based approach fails to align with proportionality rests on the assumption that established administrative-law principles apply to AI systems without material modification. This assumption is load-bearing yet receives limited justification; a dedicated subsection contrasting AI-specific features (e.g., opacity, continuous learning) with traditional administrative acts would strengthen the argument.
minor comments (2)
  1. [Abstract and Introduction] The abstract and introduction would benefit from an explicit enumeration of the specific interpretative strategies proposed, rather than leaving them implicit until later sections.
  2. [Analysis of regulatory obligations] Ensure that references to specific AI Act articles (e.g., those governing high-risk obligations) are consistently cross-referenced when discussing public-sector deployer duties.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive comment and the overall positive evaluation. The suggestion to strengthen the justification for applying administrative-law principles to AI systems is well-taken and will be addressed through a targeted revision.

read point-by-point responses

  1. Referee: [Section analysing proportionality and risk classification] The central claim that the AI Act's risk-based approach fails to align with proportionality rests on the assumption that established administrative-law principles apply to AI systems without material modification. This assumption is load-bearing yet receives limited justification; a dedicated subsection contrasting AI-specific features (e.g., opacity, continuous learning) with traditional administrative acts would strengthen the argument.

    Authors: We agree that the assumption merits explicit elaboration. In the revised version we will insert a new subsection immediately preceding the proportionality analysis. It will contrast AI-specific characteristics—opacity of decision logic, capacity for continuous learning and adaptation, scalability across large populations, and the absence of a single identifiable human decision-maker—with the features of traditional administrative acts. The subsection will explain why core principles such as proportionality and reviewability remain applicable, albeit requiring interpretive adaptation, thereby providing the missing justification for the central claim. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper advances a normative doctrinal analysis of the EU AI Act's alignment with administrative law principles such as proportionality, discretion, and the duty to state reasons. Its central claims rest on direct interpretation of the external statutory text of the AI Act and longstanding public-law doctrines, without any self-definitional loops, fitted parameters renamed as predictions, or load-bearing self-citations that reduce the argument to unverified prior work by the same authors. No equations, empirical fits, or uniqueness theorems are invoked; the proposed interpretative strategies and safeguards are presented as policy recommendations grounded in external legal sources rather than internal construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The analysis rests on standard legal axioms about the applicability of administrative law to automated decisions; no free parameters or invented entities are introduced.

axioms (1)
  • domain assumption Administrative law principles of discretion, duty to state reasons, and proportionality apply to AI-assisted public decisions in the same manner as to human decisions.
    Invoked as the baseline for evaluating the AI Act's alignment in public-sector contexts.

pith-pipeline@v0.9.0 · 5442 in / 1146 out tokens · 29525 ms · 2026-05-15T01:06:34.259907+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

9 extracted references · 9 canonical work pages

  1. [1]

    Public administration challenges in the world of AI and bots

    Agarwal, Pankaj K. “Public administration challenges in the world of AI and bots. ” Public Administration Review 78.6 (2018): 917-921. Alon-Barkat, Saar, and Madalina Busuioc. “Human–AI interactions in public sector decision making:“automation bias” and “selective adherence” to algorithmic advice. ” Journal of Public Administration Research and /T_heory 3...

    work page 2018

  2. [2]

    Humpty dumpty and high-risk AI systems: the ratione materiae dimension of the proposal for an EU arti/f_icial intelligence act

    De Cooman, Jerome. “Humpty dumpty and high-risk AI systems: the ratione materiae dimension of the proposal for an EU arti/f_icial intelligence act. ” Mkt. & Competition L. Rev. 6 (2022):

    work page 2022

  3. [3]

    Data Protection Impact Assessment: A tool for accountability and the unclari/f_ied concept of ‘high risk’ in the General Data Protection Regulation

    Demetzou, Katerina. “Data Protection Impact Assessment: A tool for accountability and the unclari/f_ied concept of ‘high risk’ in the General Data Protection Regulation. ” Computer Law & Security Review 35.6 (2019): 105342. Edwards, Lilian, and Michael Veale. “Slave to the Algorithm? Why a ‘Right to an Explanation’ is Probably Not the Remedy You Are Looki...

    work page 2019

  4. [4]

    What is the A-level algorithm? How the Ofqual’s grade calculation worked – and its effect on 2020 results explained

    Hughes, David (17 August 2020). “What is the A-level algorithm? How the Ofqual’s grade calculation worked – and its effect on 2020 results explained”. i News. Archived from the original on 21 August 2020 Geburczyk, Felix. “Automated Administrative Decision-Making under the In/f_luence of the GDPR – Early Re/f_lections and Upcoming Challenges. ” Computer La...

    work page 2020

  5. [5]

    Journal of Ethics and Legal Technologies – Volume 7(2) – November 2025 78 Lazaratos, Panagiotis

    39-62. Journal of Ethics and Legal Technologies – Volume 7(2) – November 2025 78 Lazaratos, Panagiotis. Rechtliche Auswirkungen der Verwaltungsautomation auf das Verwaltungsverfahren. Tübinger Schri&en zum Staats- und Verwaltungsrecht, Band

    work page 2025

  6. [6]

    Algorithmic audits of algorithms, and the law

    18-29. Le Merrer, Erwan, Ronan Pons, and Gilles Trédan. “Algorithmic audits of algorithms, and the law. ” AI and Ethics 4.4 (2024): 1365-1375. Mantelero, Alessandro. “/T_he Fundamental Rights Impact Assessment (FRIA) in the AI Act: Roots, legal obligations and key elements for a model template. ” Computer Law & Security Review 54 (2024): 106020. Meneceur,...

    work page 2024

  7. [7]

    Revolutionizing government operations: /T_he impact of arti/f_icial intelligence in public administration

    117-145. Mishra, Aman Kumar, and others. “Revolutionizing government operations: /T_he impact of arti/f_icial intelligence in public administration. ” Conversational Arti/f_icial Intelligence (2024): 607-634. Nnawuchi, Uchenna, and Carlisle George. “A grand entrance without a blueprint: a critical analysis of the Right to Explanation in Article 86 of the ...

    work page 2024

  8. [8]

    /T_he dynamics of AI capability and its in/f_luence on public value creation of AI within public administration

    Van Noordt, Colin, and Luca Tangi. “/T_he dynamics of AI capability and its in/f_luence on public value creation of AI within public administration. ” Government Information (arterly 40.4 (2023): 101860. Warthon, Margaret. “Restricting access to AI decision-making in the public interest: /T_he justi/f_icatory role of proportionality and its balancing fact...

    work page 2023

  9. [9]

    Holding AI-based systems accountable in the public sector: A systematic review

    179-197. Yuan, Qianli, and Tzuhao Chen. “Holding AI-based systems accountable in the public sector: A systematic review. ” Public Performance & Management Review (2025): 1-34

    work page 2025